papadopoulos1993dynamic

Dynamic Singularities in Free-Floating Space Manipulators

Authors: Papadopoulos, Dubowsky · Year: 1993 · Venue: ASME Journal of Dynamic Systems, Measurement, and Control
Raw: md

Summary

This paper introduces the concept of dynamic singularities for free-floating space manipulator systems — configurations in which the system Jacobian J* loses rank and the end-effector cannot be moved in some inertial direction, even though the kinematic structure alone would not predict a singularity. Unlike fixed-base singularities, dynamic singularities depend on the system’s mass and inertia distribution, and their occurrence at a given inertial workspace location is path dependent because spacecraft attitude is a non-integrable function of the manipulator’s motion history. The authors construct J* compactly via barycentric vectors, prove its singular structure, and partition the reachable workspace into a Path-Independent Workspace (PIW) and a Path-Dependent Workspace (PDW).

Key Claims

• For a free-floating system (no base actuation, zero external force/torque, zero initial linear and angular momentum, base CM chosen as inertial origin), the end-effector inertial linear and angular velocities can be expressed solely as functions of the controlled joint rates, via a Jacobian J* that depends on the system’s dynamics, not kinematics alone (Eq. 1, Eq. 35).
• Dynamic singularities occur where det[⁰J*(q)] = 0 (Eq. 37). These are functions of the joint configuration q with respect to the spacecraft, not of spacecraft attitude.
• Singular joint configurations cannot be mapped to unique inertial workspace points: end-effector pose depends on both q and the spacecraft attitude (e, n), and attitude is path-dependent because angular momentum is non-integrable (Eq. 38, Eq. 33c). A closed end-effector path in inertial space generally does not return the spacecraft/manipulator to the initial configuration.
• The reachable workspace splits into the PIW (no dynamic singularity for any path) and PDW (singularity may occur depending on path). A point in the PDW unreachable by a straight line may still be reached via a different path (demonstrated: straight path AB fails; path ACB with a square loop around C succeeds).
• As spacecraft mass and inertia → ∞ (m₀, I₀ → ∞), J* → J, the ordinary fixed-base Jacobian, with no change in matrix size, and dynamic singularities reduce to kinematic ones.
• Mitigations for shrinking the PDW: control spacecraft attitude (ω₀ = 0, PIW maximal, but costs fuel/complexity); use manipulator redundancy; make the spacecraft inertia large; or, for the planar case, mount the manipulator at the spacecraft CM (⁰r₀* = α = 0), which eliminates the PDW entirely.

Method

Model: an N-DOF revolute, open-chain manipulator (bodies k = 1..N) on a spacecraft base (body 0); total 6 + N DOF. Regime is free-FLOATING — the spacecraft translates and rotates passively in reaction to manipulator motion, with no attitude or position control. This is explicitly contrasted with reaction-jet-compensated and attitude-controlled categories. (For our free-FLYING, fully-actuated 6-DOF base, ω₀ is a commanded input rather than a momentum-conserving dependent variable; the dynamic-singularity mechanism here is precisely what active base control removes — see Relevance.)

Key construction:

• Barycenters. Body-fixed barycentric vectors v_ik (Eq. 11–12) compactly express each link CM position ρ_k. A striking property (Eq. 11): the inertial CM of link k depends on the positions of all links — including those after k — and on the base, unlike fixed-base chains where a link depends only on previous links.
• Momentum conservation. With p = 0 (Eq. 15) the CM is fixed (chosen as inertial origin). Angular momentum h = h_cm is built from inertia dyadics D_ijk (Eq. 21) / D_ij (Eq. 23). Eq. 22 (h_cm) is non-integrable except for N = 1 and must be carried along.
• Jacobian assembly. In spacecraft-frame matrix form (Eq. 33a–c), end-effector velocity = joint contribution + base-reaction contribution; Eq. 33c is the angular-momentum constraint used to eliminate the uncontrolled ⁰ω₀, yielding J* (Eq. 35). ⁰J₁₁ is skew-symmetric (3×3) from the CM→end-effector vector; ⁰J₁₂ (3×N), ⁰J₂₂ correspond to the end-effector Virtual Manipulator Jacobian with the first link fixed. ⁰D is the 3×3 system inertia at the CM, ⁰D_q (3×N) the inertia of moving parts.
• J* and ⁰J* are 6×N. For N = 6, J* is square and invertible when nonsingular; diag(T₀, T₀) is always nonsingular, so all singularity comes from ⁰J*(q).

Notation note: Eq. 30a is mislabeled in text as “scalar form of Eq. (19)” — it is the matrix form of Eq. (18) (the angular-velocity recursion). OCR also renders N as “7V”/“TV” in places (e.g., lines 37, 39); read as N.

Relevance to thesis

This is the canonical reference defining the Generalized/free-floating Jacobian (after Umetani & Yoshida) and the dynamic-singularity phenomenon. For a free-FLYING manipulator, the entire pathology stems from ω₀ being passively determined by Eq. 33c; with a fully actuated base, ω₀ (and r_cm) are controlled, the momentum constraint is replaced by base actuation, and J* recovers a kinematic-style structure whose singularities are no longer mass-distribution-dependent. The PIW/PDW partition and path-dependence are the precise costs a free-flying base buys out — making this paper the baseline against which free-flying coordinated control should be motivated and compared. The mitigations (CM-mounting, large base inertia, redundancy, attitude control) map directly onto design/guidance trade-offs in the risk-aware layer.

Connections

Topics: dynamic_singularity · generalized_jacobian · free_flying_vs_free_floating · path_dependent_workspace

Key Equations / Quotes

“At a dynamic singularity the manipulator is unable to move its end-effector in some inertial direction; thus dynamic singularities must be considered in the design, planning, and control of free-floating space manipulator systems.” (Abstract, p. 1)

End-effector velocity relation (Eq. 1):
$[{\dot{\mathbf{r}}}_E,{\omega }_E{]}^T={\mathbf{J}}^{\ast }\dot{\mathbf{q}}$

Free-floating Jacobian (Eq. 35a/c):
${\mathbf{J}}^{\ast }(\mathbf{e},n,\mathbf{q})=\mathrm{diag}({\mathbf{T}}_0,{\mathbf{T}}_0)\left[\begin{array}{c}-{}^0{\mathbf{J}}_{11}{}^0{\mathbf{D}}^{-1}{}^0{\mathbf{D}}_q+{}^0{\mathbf{J}}_{12}\\ -{}^0{\mathbf{D}}^{-1}{}^0{\mathbf{D}}_q+{}^0{\mathbf{J}}_{22}\end{array}\right]$

Dynamic singularity condition (Eq. 37):
$\det [{}^0{\mathbf{J}}^{\ast }(\mathbf{q})]=0$

Angular-momentum constraint eliminating the uncontrolled base rate (Eq. 33c):
$0={}^0\mathbf{D}{}^0{\omega }_0+{}^0{\mathbf{D}}_q\dot{\mathbf{q}}$

Planar two-link singular locus (Eq. 41):
$\alpha \beta D_2\sin q_1+\beta \gamma D_0\sin q_2-\alpha \gamma D_1\sin (q_1+q_2)=0$

“if the spacecraft of a space manipulator system is not actively controlled but is free-floating, then dynamic singularities can occur. All resolved rate or resolved acceleration control schemes will fail because at these points, Eq. (35) has no inverse.” (Sec. III, p. 4)

Open Questions

• Efficient construction of paths to reach PDW points: “The efficient construction of paths to reach points in the PDW is still an open area of research.” (Sec. V).
• How to exploit manipulator redundancy systematically to escape dynamic singularities (“an area which requires additional research,” Sec. VI).
• General (non-planar, 3-D) design rules for maximizing the PIW beyond the planar CM-mounting result.